Activation Tags:
@failleafMore info:
Version 2.0: Updated the dataset to my Anima Base outputs from the old Anima Preview dataset ⚠This version of the lora seems to have more issues with forgetting and I have been unable to identify the root of the issue, lower the strength to 0.7 or use more detailed prompts if you run into issues⚠
Lora trained on my outputs for people who asked for my style a while back, the dataset isn't super varied because I mostly just gen 1girl vtuber pics, it might also not be super consistent since it was trained on pictures made with different models and my style lora themselves carry a decent bit of style themselves.
I suggest lowering the strength to 0.6-0.7 and adding a few style tags to avoid potential issues that a lora trained on a pretty narrow fully synthetic dataset might have.
The style also defaults to glossy skin so you might want to put "shiny skin" in the negatives
Description
Updated from the Anima Preview dataset to an Anima Base dataset
FAQ
Comments (1)
This LoRA has been my favorite so far. Thanks so much for making and sharing it! I'm honestly amazed by every single image you've generated and posted. I keep coming back to your gallery just to admire them.
If you were to use it without your character LoRA, what settings would you go with? 1.0 or 0.7 with style tags? Which style tags do you recommend? Any other usage tips? Tysm!